Publications & Articles

Scaling Smart Robots: Combining Foundational Models with SRLs for Explainability & Customization

On using Semantic Rule Languages (SRLs) to keep advanced models explainable and customizable for real deployments. Explores how to maintain transparency and control when integrating foundation models into robotic systems.

Advisors, Not Drivers: Rethinking the Role of Foundational Models in Robotics

Argues for LLMs/VLMs as policy advisors rather than direct controllers. Discusses the importance of maintaining symbolic control structures while leveraging the power of large language models.

Building Resilience in AI-in-the-Loop Robotic Systems

A framework for dependable autonomy for real-world robots. Covers graceful degradation, human-in-the-loop recovery, and fleet-level monitoring strategies for production robotic systems.

The Role of Infrastructure in Enabling AI in the Real World

Explores the critical infrastructure components needed to deploy AI systems at scale in real-world environments. Discusses compute, connectivity, and operational considerations.

Neurosymbolic AI for Autonomous Robotics

Exploring the intersection of neural and symbolic approaches in building explainable autonomous systems.

Fleet Management at Scale

Lessons learned from managing multiple autonomous delivery robots simultaneously across different environments.

Beyond Rewards: Learning with Richer Supervision

Korupolu V. N. Pradyot, B. Ravindran

Introduced instruction-driven generalization in RL. Motivated instruction models alongside reinforcement learners as a way to accelerate learning with structured human guidance.

Instructing a Reinforcement Learner

Korupolu V. N. Pradyot, M. Sivamurugan, B. Ravindran

Formalized π-Instructions (action guidance) and Φ-Instructions (state abstraction). Demonstrated learning speedups in structured RL domains through effective combination of supervised signals with RL.

Integrating Human Instructions and Reinforcement Learners: An SRL Approach

Korupolu V. N. Pradyot, M. Sivamurugan, B. Ravindran, S. Natarajan

Used Markov Logic Networks to interpret ambiguous human instructions. Introduced probabilistic reasoning over instruction interpretations and addressed relational generalization.